1. If an idea sounds promising, I try to forget about it for a while (say a month) and if it's still a good idea after the wait, then I go for it.
Usually this means that I really have that problem and it needs fixing. So I build it.
If not, I've only lost a few brain cells. No harm done.
2. I launch, and if it works and gets people excited, then I'm really onto something.
If not, shut it down.
Note that these checkpoints allow you to back track or drop the idea. No need to keep flogging a dead horse.